Message boards :
Number crunching :
Short deadlines-High priority-suspend Milkyway
Message board moderation
Previous · 1 · 2 · 3
Author | Message |
---|---|
Send message Joined: 23 Jun 08 Posts: 53 Credit: 116,163 RAC: 0 |
Dumb Grrl Suggestion to the thread's original problem (and if someone already suggested this, sorry.) I keep all of my projects on “No New Tasks†mode. I open the valves on one, get a download of work (I maintain a 2 day cache) and I close the valves. Then, when I am on the last 2 WU’s (dual core machine) for that project, I open the floodgates on the next project. A little more labour intensive, I guess, but I crunch what I want, when I want. :) Help feed the world's hungry: Free Rice |
Send message Joined: 29 Aug 07 Posts: 486 Credit: 576,548,171 RAC: 468 |
Dumb Grrl Suggestion to the thread's original problem (and if someone already suggested this, sorry.) It's not dumb but what works for you kenzieB, I have my little ways of making projects behave nicely too. I was thinking about it this morning, there are close to 1,500,000 BOINC Participants, granted not all of them are active though. But anyway at times it seems like each one of them expects each project to cater to their individual needs instead of them learning more about how the BOINC Manager and the settings in their account really work. If more of them did that they would be able to figure it out on their own instead of the Projects having to flip flop on how they run thing because of a small minority wanting it their way ... :) |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
Alternatively, someone could attempt to convince the Burger King Corporation to form BK@Home... ;-) |
Send message Joined: 8 Nov 07 Posts: 20 Credit: 257,763 RAC: 0 |
Try setting your "Connect about every xx days" and "Additional work buffer" to very low values (mine are 0 and .05) until you've finished a longer WU. This way you should only get 1 WU at a time. But you do need to let one of the longer WUs finish to get the DCF in the ballpark. I'm not going to diddle around with my preferences that work fine everywhere else to cater to one project that can't play nice with others. BOINC should adjust. What are the run times for those three types (ballpark average)? |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
This is another thing that Poorboy can note that gets hampered by hiding computers - quick research to answer someone's question... :-( The answer, Angus, is going to be "it depends". I have no idea what kind of computer(s) you have, but my AMD FX-57 equivalent (overclocked 3700+) takes about: 317 seconds for the short tasks (gs_371) 10,025 seconds for the medium tasks (gs_373) 18,725 seconds for the longer tasks (gs_372) In the latest pull of 14 tasks, had I pulled a full set of longer tasks, that would be: 18725 * 14 / 86400 = 3.034 days My cache is set for 3 days. That's pretty tight there, and I can only complete a full set of the longest tasks if I run 24 hours a day. I actually only got 4 of the gs_372 tasks, 5 of the gs_373 tasks, and 5 of the gs_371, so in theory: ((18725 * 4) + (10025 * 5) + (317 * 5)) / 86400 = 1.465 days of actual CPU time needed, or only a 50% duty cycle... |
Send message Joined: 29 Aug 07 Posts: 486 Credit: 576,548,171 RAC: 468 |
This is another thing that Poorboy can note that gets hampered by hiding computers - quick research to answer someone's question... :-( Huh, funny, I didn't know it hampered me any Brian ... :) |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
This is another thing that Poorboy can note that gets hampered by hiding computers - quick research to answer someone's question... :-( Well, now you know, and knowing is half the battle... Yo, Joe! |
Send message Joined: 29 Aug 07 Posts: 486 Credit: 576,548,171 RAC: 468 |
This is another thing that Poorboy can note that gets hampered by hiding computers - quick research to answer someone's question... :-( Yup, but I'll still keep mine Hidden Thank You Very Much ... :) |
Send message Joined: 13 Mar 08 Posts: 804 Credit: 26,380,161 RAC: 0 |
14 WU's all due on 9/9 around 10:17pm each running 6.5 hours on my dualcore--not gonna make it |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
14 WU's all due on 9/9 around 10:17pm each running 6.5 hours on my dualcore--not gonna make it All 14 are of the gs_372 variety...? Remember that not all tasks run the same length. Here is what I posted only 5 posts down from this post to you: The answer, Angus, is going to be "it depends". I have no idea what kind of computer(s) you have, but my AMD FX-57 equivalent (overclocked 3700+) takes about: |
Send message Joined: 13 Mar 08 Posts: 804 Credit: 26,380,161 RAC: 0 |
We'll see...I'm going to let it run overnight tonight Several different #'s on the WU's...nnly a cpl 3721's |
Send message Joined: 11 Sep 08 Posts: 22 Credit: 9,081,761 RAC: 0 |
Apparently the server-supplied estimates weren’t adjusted to match; such is life in a test project … BOINC adapts to events like this over time, by adjusting the host’s Duration Correction Factor for the project, and its averages are biased to avoid downloading more work than it can handle. Actually, as one who just connected and has no project debt, such was not the case. Due to the non-adjustment of the completion times server side, and being a new project in BOINC with no prior history to base things off of, it's estimated for an 8 min completion time, and 20 WUs were grabed all in < 1 min; pretty much on a go. Thing is, it's looking like 12 hours or more to complete the first, albeit I'm basing this on 26.6% in just under 3.5 hr; I had to manually suspend all other projects and it will still expire on most of the WUs. Yes it takes time to sort itself out; but such is life when the initial estimates are off by a factor > 10:1. That's of course assuming the running WU is representative and others won't exit early. Current rate, 6 will complete; unless there's a lot of early abortions, or an optimized client one can throw on to make it complete those faster. Over time it does adjust, though slowly on each crunched WU (yes, I've crunched LHC and some projects where early abortions aren't always that uncommon and the completion times from 1 unit to another can very wildly); but the manager liked to inch things, not assume the new WU length all at once. It does mean the manager will take awhile with over deadline units; if the initial value it is given, veries from estimated minutes, and really > 10 hrs... Now in the interum set "no new work" till it gets through these or the remainder expire, but rather expect it won't have hit correct estimate before next 3 day pull, even keeping all else suspended. |
Send message Joined: 11 Mar 08 Posts: 28 Credit: 818,194 RAC: 0 |
The work units complete very quickly after they reach 50%. I would guess that the one you mention will finish in about 7 hours. As a rough guide, there are three types of work units, which you can indetify by the first three digits of the work unit number: 371 - these are the shortest 372 - these are the longest 373 - these are somewhere in the middle
|
Send message Joined: 11 Sep 08 Posts: 22 Credit: 9,081,761 RAC: 0 |
It finished in 6.5 hours in the end, it was a 372... Might be more hopeful though there's also several 373s Time will tell. The initial estimate did through it off by a lot though. |
Send message Joined: 26 Dec 07 Posts: 41 Credit: 2,582,082 RAC: 0 |
It finished in 6.5 hours in the end, it was a 372... Might be more hopeful though there's also several 373s Time will tell. The initial estimate did through it off by a lot though. On your computer, if a 372 takes 6.5 hours then I estimate a 373 will take 3.5 hours and a 371 will take just under 7 minutes. |
Send message Joined: 22 Mar 08 Posts: 90 Credit: 501,728 RAC: 0 |
It finished in 6.5 hours in the end, it was a 372... Might be more hopeful though there's also several 373s Time will tell. The initial estimate did through it off by a lot though. If you watch your manager, it will run to 50% then the wu will jump 1% per sec. to 100%. So when a wu says its at 25% its actually at 50% on the 372 and 373's. A clear conscience is usually the sign of a bad memory |
©2024 Astroinformatics Group